Documents

When work produces something worth keeping, it lands in the Library as a document. Reports, drafts, research, plans, and uploaded files use the same durable record, whether a person writes them directly, an agent creates them during a task, or a workflow produces them as an output.

Documents keep a complete history

A document has one stable identity and a sequence of immutable versions. Editing does not overwrite the previous body. It appends a version, records who made the change, and keeps the earlier state available for review or restoration. References point to the document itself, so links remain valid as its title, location, or content changes.

Markdown documents store their body directly. File-backed documents keep their content in protected storage while the Library record holds the metadata needed to find and authorize it. Both follow the same visibility and version rules.

The Library organizes shared knowledge

Documents live in folder-like paths that carry workspace permissions. A workspace can keep general material in /Library and place project-specific knowledge under a project path. Moving or renaming a document changes where people find it without changing the record that references resolve.

Archiving a folder hides that branch and its documents from Library lists, search, reference pickers, and the knowledge context agents receive. Nothing is destroyed. Restoring the folder makes the same records visible again.

Generated documents retain their source

A document created during work records the task, agent run, workflow, or playbook installation that produced it when that source exists. The source stays attached as the document is refined, so a reader can tell where the result came from without reconstructing it from a transcript.

People and agents work on the same document rather than maintaining separate human and generated record types. A draft can start in Chat, continue during a task, and become curated workspace knowledge while preserving one history.

Agents create and update documents through tama

Inside an agent run, tama documents list, search, and show retrieve visible documents. tama documents create reads a markdown body from standard input and saves a new document. tama documents update appends a version to an existing document. The current run fixes the workspace and acting agent, and the server applies the same path permissions as the app.

tama documents create --title "Customer interview synthesis" --content-path /Library/Research < synthesis.md

tama documents update <document-id> < revised-synthesis.md

A document created from task work can include --task. Task Machine uses the current agent run automatically when available. File-backed creation also accepts content type, checksum, and a server-owned storage reference through the supported upload flow.

Documents become more useful when they connect to the work around them. See References for the links that tie documents to tasks, goals, people, agents, and other workspace knowledge.